This application claims priority from Italian patent application No. MI2002A000459, filed Mar. 6, 2002, which is incorporated herein by reference.
The present invention refers to an analog-to-digital (ADC) or digital-to-analog (DAC) sigma-delta-type converter and to a method to carry out a sigma-delta-type conversion; it refers particularly to a randomizer for a sigma-delta-type converter.
In the last few years, high-resolution digital-to-analog sigma-delta-type (xcexa3xcex94) converters (DAC) have become very popular in audio applications. In particular, in order to reduce the silicon area and power consumption, multi-bit lower-order modulators are largely used. But they have however the drawback to have to use a random-number generating structure (randomizer) to interface the quantizer output levels with the following reconstruction filter. In fact, the reconstruction filter if realized with switching capacitors requires sampling capacitors whose values are subject to errors related to the technological process used to manufacture the converter, and these errors heavily influence the converter performance. However, the cumulative error, due to the presence of the capacitors, becomes equal to zero when all the capacitors have been selected for the same number of times.
Different types of randomizers are known, such as, for example, that described in the article by Rex T. Baird and Terri S. Fiez, xe2x80x9cLinearity Enhancement of Multibit DS A/D and D/A Converters Using Weighted Averagingxe2x80x9d, IEEE Trans. On Circuits and Systems, vol.42, No. 12, December 1995. This article is one of the best from the point of view of the simplicity of the circuit realization, because only one circular memory supplied by the digital output of the modulator is enough to determine which capacitors will be used at every synchronization cycle.
One of the most important limitations of this kind of structure is the fact that it does not have a good performance if applied to structures requiring an odd number of quantization levels. On the other hand (see I. Fujimori, T. Sugimoto, xe2x80x9cA 1.5 V, 4.1 mW Dual-Channel Audio Delta-Sigma D/A Converterxe2x80x9d, IEEE Journal Of Solid-state Circuits, vol. 33, No. 12, December 1998) it is important to use an odd number of levels of the quantizer to avoid the free tones produced by the low-order modulators.
Another problem is the fact that, when a signal is applied to the modulator having a low dynamic, the circular memory saturates easily.
In fact, for example in the case of a circular memory with 12 elements and with a low-level input signal, a logical word having 6 consecutive identical logic levels will frequently be used. In the presence of a high number of 6 consecutive values, the action of the randomizer will not be efficient because every two synchronization cycles the same capacitors of the reconstruction filter will be selected, nullifying the presence of the randomizer. In such a situation, tones at a frequency equal to half the synchronization frequency will be produced that will disturb the signal to be converted.
For this reason, a second-order randomizer is often used. But as is known, this often increases the complexity of the circuitry, the area occupied by the circuitry, and the power dissipated by the circuitry.
In view of the state of the art described, an embodiment of the present invention provides a sigma-delta converter that does not have the drawbacks of the known art.
According to this embodiment, the sigma-delta-type converter comprises: a sigma-delta modulator having a digital output having a first prefixed number of bits; a randomizer including a circular memory; an analogic reconstruction filter comprising a branch number equal to said first prefixed number and including sampling capacitors and a low-pass filter; characterized in that said circular memory comprises a number of elements equal to said first prefixed number of bits less one and receives at an input said first prefixed number of bits less one, and in that a bit of said first prefixed number of bits is applied to one of said branches of said reconstruction filter.
According to another embodiment of the invention, a method for carrying out a sigma-delta-type conversion comprises the steps of providing at the output of a multi-bit sigma-delta modulator a prefixed number of bits; applying at a randomizer a number of bits equal to said prefixed number of bits less one; applying the outputs of said randomizer to corresponding inputs of a converter with capacitors; applying a bit provided by said sigma-delta modulator to an input of said converter with capacitors.
Thanks to these embodiments, it is possible to avoid the saturation of the circular memory, to obtain a reduction of the quantization noise, and to avoid the generation of free tones. It is possible to use a sigma-delta modulator of low order and/or at a low frequency of operation. Furthermore, a reduction of the circuit area and a reduction of the dissipated power are obtained.